Struct StructuralTimeSeries 

pub struct StructuralTimeSeries;

Structural Time Series builder and fitter.

Implementations

impl StructuralTimeSeries

pub fn fit_local_level( y: &[f64], level_var: f64, obs_var: f64, init_level: Option<f64>, init_var: Option<f64>, ) -> StatsResult<StsFitResult>

Fit a local level model to a univariate time series.

The local level model is:

y_t = μ_t + ε_t,  ε_t ~ N(0, σ²_ε)
μ_{t+1} = μ_t + η_t,  η_t ~ N(0, σ²_η)

State x_t = [μ_t].

Arguments

• y – univariate time series (length T).
• level_var – process noise variance σ²_η (must be > 0).
• obs_var – observation noise variance σ²_ε (must be > 0).
• init_level – initial level estimate (defaults to y[0]).
• init_var – initial state variance (defaults to 1e6 for diffuse init).

pub fn fit_local_linear_trend( y: &[f64], level_var: f64, slope_var: f64, obs_var: f64, init_level: Option<f64>, init_slope: Option<f64>, init_var: Option<f64>, ) -> StatsResult<StsFitResult>

Fit a local linear trend model to a univariate time series.

The local linear trend model is:

y_t     = μ_t + ε_t,       ε_t  ~ N(0, σ²_ε)
μ_{t+1} = μ_t + ν_t + η_t, η_t  ~ N(0, σ²_η)
ν_{t+1} = ν_t + ζ_t,       ζ_t  ~ N(0, σ²_ζ)

State x_t = [μ_t, ν_t].

Arguments

• y – univariate time series.
• level_var – level innovation variance σ²_η.
• slope_var – slope innovation variance σ²_ζ.
• obs_var – observation noise variance σ²_ε.
• init_level – initial level (defaults to y[0]).
• init_slope – initial slope (defaults to 0).
• init_var – initial state variance (defaults to 1e6).